Workflow Engines

Workflow Engines are runtimes for defining, coordinating, and operating multi-step workflows across time.

They commonly provide workflow identity, workflow state or history, timers, activity scheduling, retries, compensation, signals, queries, inspection, and operational controls. Many workflow engines also provide durable execution, but durable execution is an architecture practice, not the definition of a workflow engine.

In the model, workflow engines often realize process graphs, coordination, durability of workflow material, recovery, durable execution, and long-lived observer behavior.

A workflow activation may realize an observer for a process step. The process itself may also be modeled as an entity-observer when it has identity, durable state/history, and rules for interpreting incoming events or signals over time.

A workflow engine may also be a durable execution engine when it persists enough execution material to resume, replay, retry, or recover coherent execution. A saga runtime, process manager, DAG runner, human-workflow system, or state-machine engine may be a workflow engine without exposing the same durable execution semantics.

Workflow engine concerns include:

Workflow identity.
Workflow state or history.
Durability and durable execution guarantees, if provided.
Replay, resume, or continuation behavior.
Activity execution.
Timer and signal handling.
Retry and compensation.
External interaction boundaries.

A workflow history is a persistence choice. It must still be related back to entity versions, committed events, projection models, policy scopes, and invariant scopes.